Element deletion in ABAQUS/Standard and Explicit

[chindan]

Dear all,

I am trying to carry out some impact damage and i need to do some element deletion. I ran my model using ABAQUS/Explicit and the *SHEAR FAILURE, ELEMENT DELETION = YES option and the model runs fine where by the elements which fail show zero stiffness and zero stress.

Now i need to run the model in standard since i have written a UEL describing a cohesive element. For this case i have written a UMAT to make the stress zero. The UMAT works fine when i do a uniaxial test on a single element to check my stress-strain behavior. Now when i run this material model with my impact simulation, the simulation stops prior to the element showing zero stress. I got the following error ***NOTE: THE RATE OF CONVERGENCE IS VERY SLOW. CONVERGENCE IS JUDGED UNLIKELY. Hence i reduced my time step to the minumum but i still get the error.

I am a little lost as to what are the aspects i should be checking in my simulation. In the UMAT i have gone through it wherein i know i am not dividing by any zeros. For this case i set my Jacobian as a very small value. I have written the code using the radial return method scheme and i am carrying out implicit integration.

I know the question i have posted is a little vague but i just need some guidance and some helpful tips which i should keep in mind.

Thanking you in advanced

[Jorgen]

I think your approach, in general, sounds fine. It is not easy, however, to create an effective element deletion approach in an implicit simulation. You idea to make the stress quite low is makes sense, but in practice it is different and I am not too surprised that you get convergence problems. Rapidly changing the material properties from solid to almost zero will make the stability geometry dependent and convergence problems likely.

If you tweak your Jacobian and the "zero stress" perhaps you can get convergence, if not, you are likely forced to use explicit :(

- Jorgen

[chindan]

Hi Jorgen,

Thanks for your reply. I did manage to tweak the jacobian. For this i reduce my stiffness to 1% of the original value thereby avoiding numerical instabilities. Now my simulation runs accordingly but i am facing a new problem.

The problem i face is that i have defined a sacrificial layer with these material properties, now just before failure the first element shows a stress very close to the failure stress and the final element in that layer shows a stress 10 times lower than the failure stress. Now once the first element fails it so happens that all the elements in the sacrificial layer fail and show zero stress. What is it that i am doing wrong?

Thanking you

[Jorgen]

Hmm, I don't quite understand your explanation. Can you add a picture?

- Jorgen

[chindan]

Hi Jorgen,

Thank you very much for your help. I did finally manage to solve the problem. The problem i was having was with the actual failure criterion which i was using and my implementation in the code which was incorrect. After spending numerous hours on debugging i finally managed to solve the problem.

Thanking you!